handoff prompts | PromptAura

Delegation brief that hands off a task without losing context

I need to delegate this task: task_description to delegate_name, who has this level of context already: their_existing_context. Write a delegation brief that actually transfers ownership, not just instructions. Include: the outcome I actually need (not the steps I'd take, the result), the boundary of their decision-making authority (what they can decide alone vs. what needs to come back to me), the 2-3 pieces of context I have that they probably don't, the deadline and why that deadline matters (so they can triage against other work), how and when I want a check-in (specific cadence, not "keep me posted"), and what "done" looks like in a single checkable sentence. Flag anything in the task description that sounds ambiguous enough that they'll need to ask me a clarifying question before starting.

p/agentsAauraNovice4 days agohas example⁂auto-checked

Handoff protocol between specialized agents with no context loss

Design a handoff protocol for when one specialized agent (agent_a_role) needs to pass a task to another (agent_b_role) mid-conversation. Requirements: 1. What exactly gets handed off: define the minimum sufficient package (not the full conversation dump) so agent B can pick up correctly, including what agent A already tried/ruled out so agent B doesn't repeat wasted work. 2. The trigger condition for handoff: specific signals that this is now outside agent A's competence (not a vague 'if it seems hard'), tailored to what agent A vs agent B are each actually good at. 3. What agent B is required to acknowledge before proceeding: confirm it understood the handed-off context correctly, in a form the user can also see, so a miscommunication is caught before agent B acts on a misunderstanding. 4. How the user experiences this: should the handoff be visible/announced to the user or seamless, and why, given the stakes of this specific handoff. 5. What happens if agent B determines the handoff was wrong (task actually still belongs with agent A, or belongs with a third agent) - avoid a ping-pong loop. Agent A role: agent_a_role Agent B role: agent_b_role Domain: domain

Claude GPT #multi-agent #handoff #agent-design